基于区域特征分析的层级有序码书目标检测算法

针对原码书算法在检测中码字匹配效率低以及在复杂背景下检测精确度差的问题, 提出了一种基于区域特征分析的层级有序码书目标检测新算法。首先, 利用一种新颖的“码书—码章—码字”三级结构用以精确划分码书, 在优化了码字参数之后, 将码字按访问频度进行排序并组织成定长列表。实验数据表明, 该优化策略显著地提高了算法的实时性。其次, 提出一种基于局部方向模式的运动边缘标记法, 较好地克服了噪声的影响, 并准确而完整地获取了运动物体的边缘信息。最后, 将运动边缘轮廓区域与码书区域融合, 得到了准确的运动目标。算法在时间域和空间域构建了一个能精确反映运动目标本质属性的立体模型, 经实验验证, 本算法在有随机噪声及光照变化影响的场景中, 仍有较高的识别率和较好的鲁棒性。

Detection algorithm based on regional characteristics analysis layer codebook

To solve the problem of low efficiency in the process of codeword matching and low accuracy of detection under complex environment, this paper proposed a novel layer-queue codebook target detection algorithm which was based on regional characteristics analysis. Firstly, it used a newly proposed structure of codebook-codesection-codeword to hierarchize the codebook. After optimizing the parameters of codeword, all of the codewords which were sorted according to the access frequency were organized into a fixed-length list. The experimental data indicates that the strategy has greatly improved the performance of real time. Secondly, it designed a moving-edge-marked method based on the local directional mode to obtain the edge of moving objects more accurately and it also had better robustness in noise. In the last, it fused the moving edge with the segmented movement area of codebook. This paper designed a stereoscopic model which reflected the essential attribute of the moving target in time and space according to the algorithm. The results of experiment show that the proposed method has higher recognition rate and better robustness even in natural scene with illumination variation and random noise.